Hydropower station signal source side data anti-interference acquisition method and system

Abstract

The invention discloses a hydropower station signal source side data anti-interference acquisition method and system. The method comprises the following steps: constructing a multi-source interference feature recognition network according to an environment state of a hydropower station, obtaining an interference source distribution feature spectrum, and obtaining an original signal through space-frequency domain joint filtering preprocessing based on the spectrum; performing recursion processing on the original signal by adopting an adaptive Kalman filtering algorithm to obtain a high-fidelity target signal; processing the high-fidelity target signal based on a multi-scale feature extraction and verification mechanism to obtain a credibility quantitative purification signal; and processing the credibility quantitative purification signal based on a compensation strategy to obtain a high-reliability output signal. According to the method, the influence of interference signals is gradually removed from interference source identification, filtering processing, signal reconstruction and quality evaluation to signal compensation, and finally, a result subjected to multiple anti-interference processing is obtained, so that various interferences in a complex environment of a hydropower station can be effectively resisted, and high-reliability data acquisition is realized.

Description

Technical Field

[0001] This disclosure relates to the field of data acquisition technology, and in particular to a method and system for anti-interference acquisition of data from the signal source side of a hydropower station. Background Technology

[0002] Hydropower stations, as important energy facilities, operate in extremely complex environments, including electromagnetic interference from electrical equipment and lightning interference from the natural environment. These interference signals are spatially complex and have severe spectral overlap in the frequency domain, forming a complex multi-source coupled interference environment that poses a huge challenge to the normal operation of signal acquisition systems.

[0003] Traditional anti-interference methods have significant shortcomings in dealing with the complex interference in hydropower stations. On the one hand, they lack the ability to identify interference sources, making it impossible to systematically analyze and accurately locate multi-source interference, and difficult to establish a complete interference characteristic model. On the other hand, filtering techniques have significant limitations, often employing single-domain filtering, which is insufficient to handle interference problems with complex spatial distribution and overlapping spectra. Furthermore, signal reconstruction accuracy is limited, and waveform distortion and phase shift are prone to occur under severe interference.

[0004] Existing hydropower station signal acquisition and anti-interference systems lack intelligent quality assessment mechanisms and adaptive optimization capabilities. They lack effective methods for quantitative assessment of signal quality and reliability analysis, making it impossible to provide reliable data quality assurance for subsequent systems. Furthermore, they are mostly statically configured, lacking self-learning and self-optimization capabilities, making it difficult to adapt to the dynamic changes in hydropower station operating conditions and continuously improve interference adaptability. Summary of the Invention

[0005] In view of this, the present disclosure provides a method and system for anti-interference acquisition of data from the signal source side of a hydropower station, which can solve the problems of inaccurate data acquisition by traditional methods and inability to adapt to dynamic changes in the operating conditions of hydropower stations.

[0006] This disclosure provides a method for anti-interference acquisition of data from the signal source side of a hydropower station, including: A multi-source interference feature identification network is constructed based on the environmental conditions of the hydropower station to obtain the distribution feature map of interference sources. Based on the distribution feature map of the interference source, the original signal is obtained through spatial-frequency domain joint filtering preprocessing; The original signal is recursively processed using an adaptive Kalman filter algorithm to obtain a high-fidelity target signal. The high-fidelity target signal is processed based on a multi-scale feature extraction and verification mechanism to obtain a reliable quantitative clean signal. A compensation strategy is used to process the quantified clean signal of the reliability metric to obtain a highly reliable output signal.

[0007] The anti-interference acquisition method for signal source side data of hydropower stations disclosed in this application first constructs a multi-source interference feature identification network based on the environmental conditions of the hydropower station to obtain the interference source distribution feature spectrum, thus confirming the distribution and characteristics of the interference sources. Then, based on the reliable quantitative purification signal interference source distribution feature spectrum, the original signal is obtained through spatial-frequency domain joint filtering preprocessing. This step combines information from both spatial and frequency domains, enabling more comprehensive processing of the interference signal. It considers the spatial distribution of the interference signal and effectively separates and filters overlapping spectra in the frequency domain, thereby obtaining the original signal. Next, the original signal of the reliable quantitative purification signal is processed based on an adaptive signal reconstruction algorithm to obtain a high-fidelity target signal. This step adaptively adjusts according to the actual situation of the signal, fully considering the influence of interference during signal reconstruction, and restoring the true characteristics of the signal as much as possible, thereby obtaining a high-fidelity target signal and improving the accuracy of signal reconstruction. Secondly, based on the high-fidelity target signal, a multi-scale feature extraction and verification mechanism is constructed to obtain a reliable quantitative purified signal. This step extracts and verifies the signal features at different scales, quantitatively evaluates the signal quality, and calculates the signal's reliability, thereby obtaining a reliable quantitative purified signal. Finally, the reliable quantitative purified signal is processed based on a compensation strategy to obtain a high-reliability output signal. This step dynamically compensates the signal according to changes in the hydropower station's operating conditions, further improving the signal's reliability and enabling the system to continuously improve its interference adaptability, thus obtaining a high-reliability output signal. This method gradually removes the influence of interference signals from interference source identification, filtering, signal reconstruction, quality assessment to signal compensation, effectively improving the signal's quality and reliability. The final high-reliability output signal is the result of multiple anti-interference processing steps, effectively resisting various interferences in the complex environment of the hydropower station and ensuring that the collected data is anti-interference data. Attached Figure Description

[0008] Figure 1 A flowchart illustrating the anti-interference acquisition method for signal source side data of a hydropower station provided in this embodiment of the present disclosure.

[0009] Figure 2 This is a flowchart illustrating the method for obtaining the distribution feature map of interference sources provided in this embodiment of the disclosure.

[0010] Figure 3 A flowchart illustrating the method for acquiring the original signal provided in an embodiment of this disclosure.

[0011] Figure 4 This is a flowchart illustrating a method for acquiring a high-reliability output signal provided in an embodiment of this disclosure. Detailed Implementation

[0012] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0013] Reference Figure 1 This application discloses a method for anti-interference acquisition of data from the signal source side of a hydropower station, including: S100: Construct a multi-source interference feature identification network based on the environmental conditions of the hydropower station to obtain the distribution feature map of interference sources; S200 obtains the original signal through spatial-frequency domain joint filtering preprocessing based on the interference source distribution feature map; The S300 uses an adaptive Kalman filter algorithm to recursively process the original signal to obtain a high-fidelity target signal. S400 processes high-fidelity target signals based on a multi-scale feature extraction and verification mechanism to obtain a reliable quantitative clean signal. The S500 processes the quantified clean signal based on a compensation strategy to obtain a highly reliable output signal.

[0014] This scheme addresses interference through multiple steps. Its core principle lies in the comprehensive identification, filtering, repair, and compensation of interference signals. S100 constructs a multi-source interference feature identification network to accurately identify and locate various interference sources and their characteristics, providing crucial information for subsequent processing. S200 utilizes spatial-frequency joint filtering to filter interference signals in both spatial and frequency dimensions, initially purifying the signal. S300 employs an adaptive Kalman filter algorithm to repair and enhance the original signal, improving its fidelity and quality. S400 uses a multi-scale feature extraction and verification mechanism to evaluate and screen the target signal, removing potential interference and anomalies to obtain a reliable quantitative purified signal. S500 compensates and optimizes the purified signal based on a compensation strategy, further improving its reliability and accuracy. This scheme can comprehensively and effectively combat interference in the signal source data of hydropower stations, achieving highly reliable data acquisition.

[0015] Reference Figure 2 The methods for obtaining the distribution characteristic map of S100 interference sources specifically include: S110 deploys a multi-dimensional sensor array network in hydropower stations to obtain basic data on the global electromagnetic environment.

[0016] Specifically, various types of sensors, including electric field sensors, magnetic field sensors, and current sensors, are installed in a grid-like layout in different key areas of the hydropower station, such as the generator room, transformer room, and switchyard. For example, in the generator room, an electric field sensor and a magnetic field sensor are installed at regular intervals to comprehensively monitor the electromagnetic environment of the area. These sensors are connected to a data acquisition center via wired or wireless communication. The data acquisition center collects and stores the sensor data at a certain sampling frequency (e.g., 1000 times per second) to obtain basic data on the electromagnetic environment across the entire area. This multi-dimensional sensor array network can cover all key areas of the hydropower station, comprehensively acquiring electromagnetic environment information, avoiding the limitations of single-point monitoring, and providing a rich and accurate data foundation for subsequent analysis. The real-time acquired data can reflect the dynamic changes in the electromagnetic environment, helping to promptly identify potential sources of interference.

[0017] S120 preprocesses the global electromagnetic environment basic data and extracts multidimensional features from the preprocessed global electromagnetic environment basic data.

[0018] Among them, multidimensional features include time-domain features, frequency-domain features, and time-frequency-domain features.

[0019] Specifically, the raw data is first cleaned to remove noise and outliers. For example, median filtering is used to filter electric and magnetic field data, removing abnormal data points caused by sensor errors or accidental external interference. Then, the data is normalized to unify data from different sensors to the same scale range, facilitating subsequent feature extraction and analysis. For feature extraction, in the time domain, statistical characteristics such as mean, variance, and peak value are calculated. For example, the mean of electric field data over a period reflects the average intensity of the electric field during that time period. In the frequency domain, Fourier transform is performed to convert the time-domain signal into a frequency-domain signal, extracting the frequency components and spectral features. For example, the spectrum of a magnetic field signal is analyzed to determine its main frequency components and identify any interference at specific frequencies. In the time-frequency domain, wavelet transform and other methods are used to simultaneously analyze the signal's changes in time and frequency. For example, wavelet transform is used to analyze the energy distribution of a current signal at different time and frequency scales, capturing transient changes in the signal. Preprocessing can improve data quality, remove noise and outliers, avoid the impact of these interference factors on subsequent analysis, and make feature extraction more accurate. The extraction of multidimensional features can describe the characteristics of electromagnetic environment signals from different perspectives, providing richer information for the identification and location of interference sources.

[0020] S130. Based on the historical normal operation data of hydropower station equipment, historical interference source information and their corresponding historical fault data, construct a labeled feature fingerprint database.

[0021] Specifically, data on the normal operation of hydropower station equipment over a period of time is collected, including electromagnetic environment data collected by various sensors and equipment operating parameters. Simultaneously, historical interference source information is compiled, such as the type of interference source (e.g., lightning interference, equipment failure interference), occurrence time, and location, as well as corresponding historical fault data, such as equipment damage details and electromagnetic environment characteristics at the time of the fault. Feature extraction is performed on this historical data to extract features related to the interference source, such as specific frequency electromagnetic signals and abnormal time-domain features. These features are then associated and labeled with the corresponding interference source type and fault information to construct a labeled feature fingerprint database. For example, the specific high-frequency electromagnetic signal characteristics corresponding to lightning interference and information such as the time and location of lightning occurrence are labeled and stored in the feature fingerprint database. The labeled feature fingerprint database provides a reference standard for subsequent interference source identification. By comparing the features extracted in real time with the features in the fingerprint database, the type of interference source can be quickly and accurately identified. The accumulation and analysis of historical data helps to discover potential connections between interference sources and equipment failures, providing a basis for fault prediction and prevention.

[0022] S140: The neural network is trained based on the labeled feature fingerprint database to obtain a deep learning interference source identification model.

[0023] Specifically, a suitable neural network architecture, such as a convolutional neural network (CNN) or a recurrent neural network (RNN), is selected. A training dataset is constructed based on data from a labeled feature fingerprint database, using the features in the fingerprint database as input and the corresponding interference source type as the output label. The training dataset is divided into a training set and a validation set. The neural network is trained using the training set, and the parameters of the neural network are continuously adjusted to make the model's output as close as possible to the real interference source type. During training, the performance of the model is evaluated using the validation set, and training parameters are adjusted to prevent overfitting. When the model's performance on the validation set reaches a satisfactory level, training is stopped, resulting in a deep learning interference source identification model. Deep learning models have powerful feature learning and classification capabilities, automatically learning the feature patterns of interference sources from a large amount of historical data, improving the accuracy and efficiency of interference source identification. The trained model can adapt to different types of interference sources, has good generalization ability, and can accurately identify unknown interference sources.

[0024] S150 analyzes multi-dimensional features based on a deep learning interference source identification model to obtain interference source location information.

[0025] Specifically, the multidimensional features extracted in step S120 are input into the trained deep learning interference source identification model, and the model outputs the type of interference source. Based on the type of interference source and the spatial information in the multidimensional features, combined with the geographical layout of the hydropower station and the location information of the sensors, triangulation or other positioning algorithms are used to determine the specific location of the interference source. For example, if the model identifies the interference source as equipment fault interference, and by analyzing the differences in electromagnetic signal strength collected by sensors at different locations, triangulation is used to calculate the approximate area where the interference source is located. The deep learning interference source identification model can accurately identify the type of interference source, providing an important basis for the location of the interference source. Quickly and accurately locating the interference source can help maintenance personnel take timely measures to reduce the impact of interference on hydropower station equipment and improve the operational reliability of the hydropower station.

[0026] S160. Based on the location information of the interference source, a dynamic interference propagation model is established to obtain the distribution characteristic map of the interference influence domain.

[0027] Dynamic interference propagation modeling can predict the propagation path and impact range of interference, helping operation and maintenance personnel to take protective measures in advance and reduce the impact of interference on other equipment; the interference impact domain distribution feature map can provide a reference for the optimization of equipment layout and electromagnetic protection design of hydropower stations, thereby improving the overall anti-interference capability of hydropower stations.

[0028] Furthermore, for the S160, specifically including: A100 sets boundary conditions and excitation sources based on the location information of the interference source, simulates the propagation of electromagnetic fields in the space of the hydropower station, and obtains electromagnetic field propagation simulation, which includes the distribution of electromagnetic field intensity at different locations.

[0029] First, based on the interference source location information obtained in step S150, the specific location and type of the interference source are determined. For example, if the located interference source is electromagnetic interference generated by a transformer, then in electromagnetic simulation software (such as CSTMicrowave Studio or HFSS), the location of the transformer is set as the excitation source location, and the parameters of the excitation source, such as the frequency and amplitude of the excitation source, are set according to the transformer's operating parameters (such as voltage and current). For setting boundary conditions, the actual physical boundaries of the hydropower station, such as walls and metal shells, are considered, and these boundaries are set as corresponding electromagnetic boundary conditions. For example, metal walls can be set as ideal conductor boundaries to simulate their reflection and shielding effects on the electromagnetic field. The propagation of the electromagnetic field within the hydropower station space is simulated using simulation software. The software will output the electromagnetic field intensity distribution at different locations, forming the electromagnetic field propagation simulation results.

[0030] By setting accurate boundary conditions and excitation sources, the propagation of electromagnetic fields in the complex environment of hydropower stations can be realistically simulated, providing reliable electromagnetic field data for subsequent analysis. The electromagnetic field propagation simulation results can intuitively show the distribution of electromagnetic fields in space, helping staff to quickly understand the range and intensity of electromagnetic influence of interference sources on the surrounding area, and providing a basis for electromagnetic protection.

[0031] A200: Establish a mechanical model of the hydropower station equipment, and obtain the vibration propagation path of the equipment corresponding to the location information of the interference source through finite element analysis.

[0032] For the equipment corresponding to the location information of the interference source, for example, if the interference source is a generator, a detailed 3D mechanical model of the generator is created using 3D modeling software (such as SolidWorks or ProE), including the geometry and dimensions of each component (such as rotor, stator, and casing). The created mechanical model is then imported into finite element analysis software (such as ANSYS or ABAQUS), and corresponding material properties (such as density and elastic modulus) are assigned to the model. Excitation conditions are set in the model according to the characteristics of the interference source; for example, if it is unbalanced vibration interference from the generator, a corresponding unbalanced force is applied to the rotor. Vibration analysis is then performed using finite element analysis software to calculate the propagation path of the vibration within the equipment and surrounding structures, obtaining a visualization of the vibration propagation path. Mechanical models and finite element analysis provide a deep understanding of the vibration characteristics and propagation laws of the equipment, helping to identify paths where vibration may affect other equipment or structures. By understanding the vibration propagation path, targeted vibration reduction measures can be taken to reduce vibration damage to hydropower station equipment and structures, and improve the service life and operational stability of the equipment.

[0033] A300 establishes an acoustic propagation model for hydropower station equipment. Based on the location information of the interference source, it obtains sound field distribution information, which includes the sound pressure level distribution at different locations.

[0034] Taking the equipment corresponding to the interference source as an example, an acoustic propagation model of the equipment is established using acoustic simulation software (such as COMSOL Multiphysics or LMS Virtual.Lab). This model needs to consider the geometry of the equipment, the acoustic properties of the materials (such as the sound absorption coefficient), and the spatial layout of the hydropower station. Based on the location information of the interference source, the position and parameters of the sound source are set in the model. For example, if the interference source is the noise of a water pump, the sound power level and frequency characteristics of the sound source are set according to the power and operating status of the water pump. The acoustic simulation software is used to simulate the propagation of sound in the space of the hydropower station, calculate the sound pressure level distribution at different locations, and obtain the sound field distribution information. The acoustic propagation model can accurately predict the propagation and distribution of sound in the complex space of the hydropower station, help assess the impact of noise on workers and the surrounding environment, and by understanding the sound field distribution information, effective noise reduction measures can be taken, such as reasonably arranging sound-absorbing materials and optimizing equipment layout, to improve the working environment and reduce noise pollution.

[0035] A400, based on electromagnetic field propagation simulation, vibration propagation path and sound field distribution information, uses the principle of linear superposition to add the interference intensity generated by different interference sources at the same location, resulting in a superposition intensity distribution map of a single spatial point.

[0036] Specifically, interference intensity data, such as electromagnetic field strength, vibration intensity, and sound pressure level, are extracted from electromagnetic field propagation simulations, vibration propagation paths, and sound field distribution information at the same spatial point. For different types of interference intensity data, normalization is performed according to their physical meaning to ensure comparability; for example, electromagnetic field strength, vibration intensity, and sound pressure level are converted into relative intensity values. Using the principle of linear superposition, the normalized interference intensities generated by different interference sources at the same spatial point are added together to obtain the superimposed intensity value for that spatial point. This process is repeated for all spatial points within the hydropower station area, ultimately generating a superimposed intensity distribution map for each individual spatial point. The linear superposition principle comprehensively considers the superimposed effects of different types of interference sources at the same location, providing a more complete reflection of the actual interference situation. The superimposed intensity distribution map visually displays the comprehensive interference intensity at different locations, helping staff quickly identify areas with severe interference so that targeted protective measures can be taken.

[0037] A500 obtains the distribution characteristic map of interference sources based on the superimposed intensity distribution map of all spatial points.

[0038] Further processing is performed on the superposition intensity distribution map of individual spatial points. For example, interpolation algorithms are used to smooth the data, making the distribution map more continuous and accurate. Based on the magnitude of the superposition intensity, the hydropower station space is divided into different regions; for example, areas with higher superposition intensity are marked as high-interference zones, and areas with lower superposition intensity are marked as low-interference zones. Combining the geographical layout and equipment distribution information of the hydropower station, an interference source distribution characteristic map is drawn, clearly showing the location and range of different interference areas. This interference source distribution characteristic map intuitively displays the distribution of interference within the entire hydropower station, facilitating staff to quickly understand the overall interference situation. This characteristic map provides important reference for equipment layout adjustments, electromagnetic protection design, and acoustic and vibration control of the hydropower station, contributing to optimizing the operating environment and improving equipment reliability.

[0039] In this embodiment, by comprehensively considering multiple interference factors such as electromagnetic fields, vibration, and sound fields, the impact of interference sources on hydropower station equipment and the environment can be fully assessed, avoiding the limitations of single-factor assessment. The interference source distribution characteristic map can clearly show the location and range of high-interference areas, helping operation and maintenance personnel to accurately locate areas requiring key protection and control, improving the pertinence and effectiveness of protective measures. Based on the interference source distribution characteristic map, the equipment layout can be reasonably adjusted during the hydropower station design phase to avoid mutual interference between equipment. During the operation phase, measures can be taken in a timely manner to reduce interference, improve the operating efficiency and reliability of the hydropower station, accurately understand the distribution of noise and electromagnetic interference, and help to take effective protective measures to ensure the health and safety of workers and reduce the risk of occupational diseases.

[0040] Reference Figure 3 The specific methods for acquiring the raw S200 signal include: S210 performs spatial spectrum estimation on the distribution feature map of the interference source to obtain the directional filter signal.

[0041] S220, based on the directional filtering signal, constructs an adaptive frequency domain notch filter bank to obtain a spectrum-cleaned signal.

[0042] The signal obtained through this step has achieved dual suppression in both the directional and frequency domains.

[0043] S230, based on the spectral cleaned signal, performs multi-channel correlation analysis and blind source separation to obtain the preliminarily cleaned original signal.

[0044] In this embodiment, the original signal is progressively purified through directional-frequency domain dual suppression and multi-channel processing to improve signal quality and usability. Specifically, in complex environments, signals are often affected by various interferences. Directional-frequency domain dual suppression can specifically eliminate interference from specific directions and frequencies, making the signal purer and improving the signal-to-noise ratio. The purified signal can provide more accurate information in subsequent processing. For example, in signal analysis and feature extraction, reducing the impact of interference can make the results more reliable, thereby improving the overall system performance. Through multi-channel correlation analysis and blind source separation, different source signals mixed together can be separated, making it possible to process and analyze each source signal individually.

[0045] The specific methods for obtaining the S210 "directional filter signal" include: S211. Based on the MUSIC algorithm or ESPRIT algorithm, the interference azimuth angle is extracted from the interference source distribution feature map. That is, the two algorithms MUSIC (Multiple Signal Classification) or ESPRIT (Estimation of Signal Parameters via Rotational Invariance Techniques) are used to analyze the azimuth angle of the interference source relative to the reference point from the interference source distribution feature map.

[0046] Compared with traditional azimuth estimation methods, these two algorithms have higher resolution and accuracy, and can more accurately locate the position of the interference source. The interference azimuth angle is an important parameter for subsequent directional filtering. By accurately obtaining the interference azimuth angle, the subsequent filtering operation can be more targeted, effectively suppressing interference signals from specific directions and improving signal quality.

[0047] S212 uses the interference azimuth angle as the null guide and obtains the weighting coefficients of the corresponding array elements through an adaptive beamforming algorithm.

[0048] The adaptive beamforming algorithm can automatically adjust the weighting coefficients of the array elements according to the interference azimuth angle, so that the beam forms nulls in the interference direction, thereby effectively suppressing interference signals from that direction. This adaptive characteristic enables the system to maintain good performance under different interference environments, improving the robustness of the system. By adjusting the weighting coefficients of the array elements, the beamforming algorithm can make the array form the main beam in the direction of the desired signal, enhancing the reception capability of the desired signal. At the same time, it forms nulls in the interference direction, reducing the influence of interference signals, thereby improving the signal reception quality.

[0049] S213 applies weighting coefficients to the array-received signal to obtain a directional filtered signal.

[0050] The directional filter signal is x1(t): x1(t) = wᴴ·x(t), where w is the weighting vector, wᴴ is the conjugate transpose of w, and x(t) is the array received signal of the corresponding array element.

[0051] Applying weighting coefficients to the array-received signal is equivalent to performing directional filtering. This effectively suppresses interference signals from a specific direction while preserving the desired signal. Directionally filtered signals offer higher quality and usability in subsequent signal processing, providing more accurate information for later analysis and processing. Since the directionally filtered signal has already performed initial suppression of interference signals, it reduces the complexity of subsequent processing. In subsequent signal processing, the directionally filtered signal can be directly manipulated without needing to consider interference signals from a specific direction, thus improving processing efficiency.

[0052] The method for obtaining the S220 "spectrum cleanup signal" specifically includes: S221 performs a short-time Fourier transform on the directional filtered signal to obtain a time-frequency diagram.

[0053] Specifically, a short-time Fourier transform is performed on the directional filter signal to obtain a time-frequency plot. The short-time Fourier transform is a time-frequency analysis method that segments the signal in time and performs a Fourier transform on each segment to obtain the distribution of the signal at different times and frequencies. In this way, the change of the signal's spectrum over time can be observed, which helps to analyze the interference characteristics in time-varying signals. Converting the time-domain signal (i.e., the directional filter signal) to a time-frequency plot in the time-frequency domain facilitates the subsequent identification and processing of interference in the frequency domain.

[0054] S222, Identify the location of interference spectral lines and their corresponding bandwidths from the time-frequency diagram.

[0055] Specifically, in the time-frequency diagram, the interference signal usually manifests as peaks in the spectrum. By using the peak tracking algorithm, the frequency positions corresponding to these peaks at different time points can be found, i.e., the positions of the interference spectral lines. After determining the positions of the interference spectral lines, the bandwidth of the interference signal in the spectrum is further determined so as to determine the operating range of the notch filter. Accurately locating the position and range of the interference signal in the frequency domain provides the necessary parameters for constructing an adaptive notch filter.

[0056] S223. Based on the position and bandwidth of the interference spectral lines, construct an adaptive notch filter group. The number of notch filters in the adaptive notch filter group is the same as the number of interference spectral line positions, and each notch filter corresponds to one interference signal.

[0057] Specifically, this includes: 1) Determining the number of notch filters; based on the number of interference spectral line positions obtained by identification, the number of notch filters to be constructed is determined. Each notch filter corresponds to one interference signal, so the number of notch filters is equal to the number of interference signals.

[0058] 2) Design the frequency response of a single notch filter; a common notch filter frequency response function can be implemented using a second-order IIR (Infinite Impulse Response) notch filter, whose transfer function is: ; ; Where i = 1, 2, ..., K represents the th... One notch filter; It is the first The normalized center frequency of a notch filter It is the first The center frequency of the interference spectral line It is the sampling frequency; It is a parameter related to the bandwidth of the notch filter.

[0059] 3) Determine the notch depth; the notch depth represents the notch filter's ability to suppress interference signals, and is usually adaptively adjusted based on the signal-to-interference-plus-noise ratio (SINR). Let the notch depth be... The signal-to-interference-plus-noise ratio of the interfering signals is The following method is used to determine the notch depth. :when A lower value indicates stronger interference, requiring a larger notch depth. The setting can be adjusted accordingly. ( (for maximum notch depth); when A higher value indicates weaker interference, and the notch depth can be appropriately reduced. ,in Minimum notch depth The signal-to-interference-plus-noise ratio (SINR) threshold. To achieve the maximum signal-to-interference-plus-noise ratio (SNR), in the frequency domain, multiply the frequency response of a single notch filter by the notch depth to obtain the adjusted first... Frequency response of a notch filter .

[0060] 4) Combined Notch Filter Group: Combining the frequency responses of all adjusted individual notch filters yields an adaptive notch filter group. : ,in, This represents the total number of notch filters.

[0061] In this step, the center frequency of the notch filter is dynamically adjusted according to the position of the interference spectral line, ensuring that the notch filter is always aligned with the frequency of the interference signal. The notch depth represents the notch filter's ability to suppress interference signals. It adaptively adjusts the notch depth based on the signal-to-interference-plus-noise ratio (SINR). When the SINR is low, indicating strong interference, the notch filter increases its depth to suppress the interference more effectively. When the SINR is high, the notch filter can appropriately reduce its depth to avoid unnecessary loss of the useful signal. The adaptive notch filter array can suppress interference in a targeted manner based on the real-time characteristics of the interference signal.

[0062] S224 multiplies the time-frequency graph with the adaptive notch filter group point by point, and then performs an inverse transformation on the multiplication result to obtain the spectral clean signal.

[0063] Specifically, in the frequency domain, the time-frequency graph is multiplied point-by-point with the adaptive notch filter array. Since the notch filter array has a smaller value at the interference frequency, the multiplication effectively suppresses the spectral components of the interference signal. Then, the processed time-frequency domain signal is converted back to the time domain, yielding the spectrally cleaned signal. The output spectrally cleaned signal undergoes both directional and frequency domain suppression, reducing the impact of interference to a certain extent.

[0064] The method for acquiring the raw signal in S230 specifically includes: S231, the spectrum cleansing signal is divided into frames to obtain the signal correlation matrix between channels.

[0065] Specifically, the spectral cleansing signal (i.e., a continuous signal) is divided into several short-segment frames. Each frame can then be processed individually, improving processing efficiency and accuracy. In this embodiment, the spectral cleansing signal is a multi-channel signal (e.g., a signal received from multiple sensors). Its covariance matrix is ​​then calculated. The covariance matrix reflects the correlation between signals from different channels, thus obtaining the inter-channel signal correlation matrix. Further, assuming the signal has M channels, the covariance matrix... It is an M×M matrix whose elements Indicates the first The first channel and the first The covariance between the signals of each channel is calculated using the following formula: ,in and They are the first The and the first The signal from each channel, E[] represents the mathematical expectation. The covariance matrix represents conjugation, and the correlation between signals from different channels can be analyzed.

[0066] After framing, each frame can be processed separately, avoiding processing the entire long signal all at once, reducing computational load and improving processing speed. Signal characteristics may differ at different time periods, and framing processing can better capture the local features of the signal, thus analyzing the signal more accurately. By calculating the covariance matrix, the correlation between signals in different channels can be clearly understood, providing important information for subsequent blind source separation.

[0067] S232, a blind source separation model is established based on the signal correlation matrix between channels.

[0068] Specifically, the blind source separation model includes an objective function and independent component analysis (ICA) or independent vector analysis (IVA) algorithms, with the objective function being: , ,in, As a measure of non-Gaussianity, As a measure of time-frequency continuity, These are the weighting coefficients.

[0069] Based on the inter-channel signal correlation matrix obtained in the previous step, a blind source separation model is established: First, the Independent Component Analysis (ICA) algorithm is selected (the Independent Vector Analysis (IVA) algorithm can also be selected); the objective function is determined. The non-Gaussianity metric can be calculated using methods such as kurtosis, and the time-frequency continuity metric can be measured by indices such as the continuity of the time-frequency distribution. For example, for the non-Gaussianity metric, the kurtosis value of the separated signal can be calculated. The larger the kurtosis value, the more the signal deviates from the Gaussian distribution. The blind source separation model can separate the mixed source signals based on the inter-channel correlation information, allowing us to extract each independent source signal from the mixed signal. The objective function comprehensively considers both non-Gaussianity and time-frequency continuity, ensuring that the separated signal achieves good results in both aspects while separating the signal, thus improving the separation quality.

[0070] S233, based on the blind source separation model, iteratively optimizes the separation matrix to obtain the separated signal components that meet the iterative convergence condition.

[0071] A separation matrix is ​​randomly initialized. For example, if the signal has 3 channels in step 1, and we expect to separate N=3 source signals, then the separation matrix is ​​a 3×3 matrix. Iterative optimization is performed using a gradient descent-based method, with the iterative formula as follows: ,in It is the learning rate, which can be optimally set. It is 0.01. It is the objective function Regarding the gradient of the separation matrix; in each iteration, calculate the gradient of the objective function corresponding to the current separation matrix, and then update the separation matrix; the iteration process continues until a convergence condition is met, such as the change in the objective function being less than a preset threshold (e.g., 10). −5 The resulting separation matrix is ​​then used to process the inter-channel signal correlation matrix to obtain the separated signal components. Through iterative optimization, the separation matrix is ​​continuously adjusted to make the separated signal components better meet the requirements of the objective function, thereby improving the accuracy and quality of blind source separation. The iterative process can adaptively adjust the separation matrix according to the actual signal conditions, adapting to different signal characteristics and noise environments.

[0072] S233, select the signal component that best matches the prior target features (such as modulation pattern, sparsity, spectral correlation), and denote it as the target component.

[0073] Assuming the target signal's modulation pattern is Binary Phase Shift Keying (BPSK), its sparsity is characterized by significant energy concentration in certain frequency bands, and its spectral correlation exhibits a specific periodicity. For the separated signal components obtained in the previous step, we calculate their matching degree with the prior target features. For example, for the modulation pattern, we can analyze the signal's phase change to determine if it conforms to BPSK modulation; for sparsity, we can calculate the signal's energy distribution in different frequency bands and compare it with the prior sparsity features; for spectral correlation, we can calculate the signal's spectral correlation function and match it with the prior spectral correlation periodicity. The signal component with the highest matching degree is selected as the target component. We accurately extract the desired target signal from the separated signal components, eliminating the influence of other interference signals, and selecting based on prior target features to ensure the extracted target component better meets the requirements of the practical application.

[0074] S234 performs amplitude normalization and phase correction on the target component to obtain the preliminarily purified original signal.

[0075] Specifically, for the target component, amplitude normalization is first performed, and the maximum amplitude value A of the target component is calculated. max Then divide each sample value of the target component by A. max This ensures the signal amplitude range is between [−1, 1]. Next, phase correction is performed. Given the ideal phase characteristics of the target signal, the phase error is calculated by comparing the phase of the target component with the ideal phase. The phase of the target component is then adjusted to conform to the ideal phase characteristics. Amplitude normalization provides a uniform scale for the signal amplitude, facilitating subsequent processing and analysis; phase correction eliminates potential phase deviations that may occur during signal transmission, resulting in more accurate signal phase and improved signal quality and recognizability.

[0076] The method disclosed in this embodiment can effectively separate individual source signals from a mixed spectral cleansing signal and extract the target signal we need. At the same time, the target signal is cleaned to remove noise and interference. Through a series of steps such as frame processing, blind source separation, feature matching, amplitude normalization and phase correction, the quality and accuracy of the signal are improved, providing a better foundation for subsequent signal analysis and applications. This scheme can be adjusted according to different signal characteristics and prior target characteristics to adapt to a variety of different signal processing scenarios and needs.

[0077] In the S300 algorithm, "Adaptive Kalman filtering algorithm is used to recursively process the original signal to obtain a high-fidelity target signal," the system state is optimally estimated through two steps: prediction and update. This is achieved by utilizing the system's state equation and observation equation, combined with noise statistical characteristics. The basic steps include: state prediction, prediction covariance update, Kalman gain calculation, state update, and covariance update, ultimately yielding the high-fidelity target signal.

[0078] Specifically, 1) Define the system model, including state equations and observation equations.

[0079] State equation: Assume the state of the original signal can be represented by a state vector x k The state equation describes the relationship between the state vector and time, and its general form is: x k =F k x k−1 +B k u k +w k , of which F k It is the state transition matrix, describing how the system state transitions from the previous time step (time step k-1) to the current time step (time step k). B k It is the control input matrix (this matrix controls the input vector u) k In relation to the system state, when the system is subjected to external control, B k (This determines the degree and manner in which control inputs affect the system state), u k It is the control input vector (representing the external control signal applied to the system at time k), w k It is a process noise vector (a random vector representing unmodeled uncertainties within the system), typically assumed to have a mean of zero and a covariance of Q. k The Gaussian distribution, i.e., w k ~N(0,Q k ).

[0080] Observation equations: Observation equations describe how to obtain observations from the system state, and are in the form of: zk =H k x k +v k , where z k It is the observation vector (the actual measured value obtained from observing the system state at time k), H k It is the observation matrix (which will take the system state vector x) k Mapping to the observation space describes how to obtain observations from the system state, v k It is an observation noise vector (a random vector representing the noise introduced during the observation process), which follows a pattern with zero mean and covariance R. k The Gaussian distribution, i.e., v k ~N(0,R k ).

[0081] 2) Initialize the state estimate and the estimated covariance matrix. These two values ​​can be set based on prior knowledge, and are usually set to a large value to indicate a high degree of uncertainty about the initial state. Initialize the process noise covariance matrix and the observation noise covariance matrix. The values ​​of these two matrices need to be estimated based on the actual situation, for example, through statistical analysis of historical data.

[0082] 3) Recursive processing (repeatedly performing prediction and update steps) The prediction steps include: 3.1) State Prediction: Predict the estimated state value at the current moment based on the state equation: ; 3.2) Covariance Prediction Update: Predict the estimated covariance matrix for the current time step: ,in, Let w be the process noise covariance matrix, which describes the process noise vector w. k The statistical properties of the system reflect the degree of uncertainty and interrelationships of different state components during the system's state transition process.

[0083] 4) Adaptive adjustment of noise covariance: In practical applications, the statistical characteristics of process noise and observation noise change over time. Adaptive methods are used to adjust these characteristics. and R k For example, residual sequences can be used to estimate the observation noise covariance R. k For process noise covariance It can be adjusted according to the dynamic characteristics of the system and the error of the state estimation.

[0084] 5) Update steps: 5.1) Kalman Gain Calculation: Calculate the Kalman gain based on the prediction covariance matrix and the observation noise covariance matrix. ,in, The observation noise covariance matrix describes the observation noise vector v. k The statistical characteristics reflect the magnitude and interrelationship of measurement errors of different observation components during the observation process; The covariance matrix is ​​used to predict the estimated state values. The degree of uncertainty reflects the uncertainty introduced during the prediction process due to system model and process noise. Kalman gain It is a matrix that determines the weight with which the observed values ​​should be used to correct the predicted state during state updates. It is used to balance the reliability of the predicted state and the observed values, so that the final state estimate can comprehensively consider the system model and the actual observation information to achieve the optimal estimate.

[0085] 5.2) State Update: Update the state prediction using the observed values ​​to obtain the optimal state estimate at the current time. ,in, This is the predicted state estimate. The updated state estimate. At time k, the optimal state estimate is obtained after updating the predicted state and the observed value using the Kalman filter algorithm. It is the output of the Kalman filter algorithm and represents the best estimate of the current state of the system, which can be used for subsequent analysis and decision-making.

[0086] 5.3) Covariance Update: Update the estimated covariance matrix: , representing the updated state estimate The degree of uncertainty reflects the reduction in uncertainty during the state update process due to the use of observation information, providing a basis for the prediction covariance update at the next time step, and affecting the Kalman gain calculation and state update at subsequent time steps.

[0087] After the above recursive processing, the final state estimate is obtained. This is a high-fidelity target signal.

[0088] The specific methods for obtaining the S400 reliable quantitative cleanup signal include: S410 employs wavelet transform to perform multi-scale decomposition of high-fidelity target signals. By selecting appropriate wavelet basis functions (such as Daubechies wavelets), the signal is decomposed into sub-signals of different scales and frequencies.

[0089] Suppose we have a high-fidelity audio signal as the target signal, and we choose the Daubechies wavelet (such as the commonly used db4 wavelet) as the wavelet basis function. Using a wavelet transform algorithm, we decompose the audio signal into sub-signals of different scales and frequencies. For example, we can decompose the signal into approximate sub-signals (representing the low-frequency part of the signal, i.e., the overall outline of the signal) and detail sub-signals (representing the high-frequency part of the signal, containing detailed information about the signal). Generally, we can perform 3-5 levels of decomposition to obtain sets of sub-signals at different scales. Wavelet transform has the characteristic of multi-resolution analysis, which can decompose the signal at different scales and frequencies. Through multi-scale decomposition, we can separate the low-frequency and high-frequency components of the signal, which facilitates subsequent targeted processing of different frequency components. Sub-signals at different scales contain information about the signal at different resolutions, which helps to analyze the characteristics of the signal more comprehensively.

[0090] S420 extracts the corresponding target features from the sub-signals at each scale after decomposition; the target features include one or more of the following: energy features, variance features, and mean features.

[0091] For each decomposed sub-signal, its energy characteristic, variance characteristic, and mean characteristic are calculated. Taking the energy characteristic as an example, for a certain sub-signal x(n), its energy E is expressed by the formula... The calculation involves finding the sub-signal length, where N is the sub-signal length. Variance features are obtained by calculating the sample variance of the sub-signal, while the mean feature is the average of all sample values ​​of the sub-signal. Target features (such as energy, variance, and mean) reflect the statistical characteristics and distribution of the sub-signal. Different features can describe the characteristics of the sub-signal from different perspectives. For example, energy features can represent the intensity of the sub-signal, variance features can reflect the degree of fluctuation of the sub-signal, and mean features can reflect the average level of the sub-signal. Extracting these features helps in subsequent in-depth analysis and judgment of the signal.

[0092] S430 performs feature anomaly verification on the target features and determines the confidence weight of each sub-signal based on the verification results.

[0093] First, a normal range is defined for each target feature. For example, for the energy feature, the mean energy μ is obtained through statistical analysis of a large number of normal signal samples. E and standard deviation σ E Set the normal range to [μ E −kσ E ,μ E +kσ E ]Where k is a constant; for each target feature of a sub-signal, if its value is within the normal range, the feature is considered normal; if it exceeds the normal range, the feature is considered abnormal; the confidence weight of each sub-signal is determined based on the number and severity of abnormal features; for example, if all features of a sub-signal are normal, its confidence weight is 1; if some features are abnormal, its confidence weight is appropriately reduced according to the abnormality, such as to 0.8 or 0.6. Feature anomaly verification can help us identify sub-signals that may be interfered with or have anomalies. By determining the confidence weight, the reliability of different sub-signals can be quantitatively evaluated. In subsequent signal processing, sub-signals with high confidence are given higher weights, and sub-signals with low confidence are given lower weights, thereby reducing the impact of abnormal sub-signals on the final result.

[0094] S440, based on the confidence weight, weights and combines the sub-signals of each scale to obtain the confidence-quantified clean signal.

[0095] Suppose there are M sub-signals x1, x2, ..., xn of different scales. M The corresponding credibility weights are w1, w2, ..., w M The credibility quantification of the purified signal y is then expressed through the formula. The calculations show that by using weighted combination, sub-signals of different scales and confidence levels can be processed together. Sub-signals with higher confidence levels play a greater role in the combination, while sub-signals with lower confidence levels have a relatively smaller impact. This approach can effectively suppress interference from anomalous sub-signals while preserving useful information in the signal, thereby improving the signal quality and confidence.

[0096] The methods disclosed in S410-S440, through multi-scale decomposition and feature extraction using wavelet transform, can more comprehensively analyze signal characteristics, identify potential anomalous components, and effectively suppress the influence of anomalous signals by determining and weighting confidence weights, thereby improving signal purity and reliability. This scheme can dynamically adjust confidence weights based on the actual characteristics and anomalous situations of the signal, exhibiting strong adaptability and effectively purifying signals under various signal environments and interference conditions. Quantitatively purified signals based on confidence levels have higher usability in subsequent signal processing and analysis. For example, in tasks such as target detection and recognition, purified signals can provide more accurate information, improving the accuracy and reliability of the task.

[0097] Reference Figure 4 Regarding the S500 method of "processing the quantified clean signal based on a compensation strategy to obtain a high-reliability output signal," the method for obtaining the high-reliability output signal specifically includes: S510 uses the 3σ criterion to identify outliers in the credible quantification clean signal.

[0098] Suppose we have a set of reliable quantitative purification signals, such as temperature data collected by a sensor over a period of time. First, we calculate the mean μ and standard deviation σ of this set of signals. According to the 3σ criterion, normal data should fall within the interval [μ−3σ,μ+3σ]; for each data point in the signal, if its value exceeds this interval, it is identified as an outlier. The 3σ criterion is a simple and effective outlier identification method. Based on the statistical characteristics of the data, it can accurately identify data points that deviate from the normal distribution in most cases. This method does not require complex models or a large amount of prior knowledge, has relatively low computational cost, and is easy to implement.

[0099] S520, determine the corresponding compensation strategy based on the type of outlier, denoted as the target strategy.

[0100] For identified outliers, determine their type and severity. If the outlier is small, such as a small difference between the outlier and the normal range boundary (e.g., in the temperature data corresponding to the 3σ criterion [19℃, 31℃], the outlier is 32℃, exceeding the normal range by 1℃), interpolation can be used for compensation. Taking linear interpolation as an example, a reasonable replacement value is obtained by linear calculation based on the two normal data points before and after the outlier.

[0101] For significant outliers, such as those with a large difference between the outlier and the normal range boundary (e.g., an outlier of 40℃), predictive compensation can be made based on historical data and trends. For instance, by analyzing temperature trends over a period of time, a simple linear or nonlinear model can be established, and the reasonable temperature value at that moment can be predicted as the compensation value.

[0102] Different types and degrees of outliers have different effects on signals. Different compensation strategies can be used to deal with outliers more effectively. For small-amplitude outliers, interpolation is simple and quick and can compensate without introducing too much error. For large-amplitude outliers, predictive compensation based on historical data and trends can better take into account the overall change pattern of the signal and improve the accuracy of compensation.

[0103] The S530 performs compensation based on the target strategy to obtain a highly reliable output signal.

[0104] Outliers are compensated according to the determined target strategy. If interpolation is used, the compensation value is calculated according to the interpolation formula and replaced with the compensation value. If predictive compensation is used, the compensation value is calculated based on the established predictive model and replaced with the original outlier. After compensating for all outliers, a new set of signal data is obtained, which is the high-reliability output signal. By compensating for outliers, the adverse effects of outliers on the signal can be eliminated, making the signal smoother and more stable. The high-reliability output signal can more accurately reflect the real physical quantity or phenomenon, providing a more reliable basis for subsequent analysis and decision-making.

[0105] The method disclosed in this embodiment effectively reduces the interference of abnormal data on signals by identifying and compensating for outliers, thereby improving signal quality and reliability. In practical applications, high-reliability signals can improve system performance and stability. For example, in sensor data processing, they enable more accurate monitoring and control of physical quantities. The processed high-reliability output signal has higher availability in subsequent data processing and analysis. Whether for data analysis, model training, or decision-making, reliable data improves the accuracy and credibility of the results. This scheme employs different compensation strategies based on the type and severity of outliers, exhibiting strong adaptability and applicability to different types of signals and various abnormal situations. It can effectively handle outliers and improve signal quality in multiple scenarios.

[0106] Furthermore, the operating environment of hydropower stations is complex, with multiple sources of interference, such as electromagnetic interference from electrical equipment and lightning interference from the natural environment. Constructing a multi-source interference feature identification network can comprehensively and accurately identify the characteristics of different types of interference sources. By analyzing and integrating these features, an interference source distribution feature map can be obtained, providing basic information for subsequent anti-interference processing, so as to design targeted anti-interference strategies.

[0107] Interference signals typically exhibit certain spatial and frequency distribution characteristics. Spatial-frequency joint filtering can combine the distribution characteristic spectrum of interference sources to filter signals in both spatial and frequency dimensions. Spatially, it can filter out interference signals from specific directions; in terms of frequency, it can remove interference noise from specific frequency bands, thereby initially purifying the signal and obtaining a relatively pure original signal.

[0108] The original signal after filtering preprocessing may still have some residual interference and distortion. The adaptive Kalman filter algorithm can automatically adjust parameters according to the real-time characteristics of the signal to reconstruct and repair the original signal. It can adaptively track the changes in the signal, compensate for the missing and distorted parts in the signal, thereby improving the signal quality and fidelity and obtaining a more accurate target signal.

[0109] The target signal may also contain some potential interference or anomalies. Multi-scale feature extraction can analyze the target signal at different scales and frequencies to extract richer feature information. Through the verification mechanism, the extracted features can be evaluated and screened to determine the credibility of the signal. Parts with low credibility can be processed or removed to obtain a quantified clean signal with credibility, thereby further improving the reliability of the signal.

[0110] Even after the preceding processing, the purified signal may still have some minor errors or omissions. The compensation strategy can compensate and correct the purified signal based on its historical data, statistical characteristics, and real-time changes. Through reasonable compensation, the accuracy and reliability of the signal can be further improved, ultimately resulting in a highly reliable output signal that meets the needs of practical applications.

[0111] The disclosed method for anti-interference acquisition of signal source data in hydropower stations effectively solves the problems existing in the prior art through a series of specific technical means, ensuring the anti-interference performance of the acquired data. Specifically, in S100, a multi-source interference feature identification network is constructed based on the environmental conditions of the hydropower station. Due to the complex operating environment of hydropower stations, there are multiple interference sources such as electromagnetic interference from electrical equipment and lightning interference, and the interference signals are complexly distributed in space and frequency domains. This network can systematically analyze multi-source interference, identify and extract the features of various interference sources, and thus obtain an interference source distribution feature map. This map clearly defines the distribution and characteristics of the interference sources, solving the problem that traditional methods cannot systematically analyze and accurately locate multi-source interference, and providing a foundation for subsequent anti-interference processing. S200 uses spatial-frequency domain joint filtering preprocessing based on the interference source distribution feature map. Traditional methods often employ single-domain filtering, which struggles to handle interference with complex spatial distribution and overlapping spectra. Space-frequency joint filtering, however, combines information from both spatial and frequency domains, enabling more comprehensive processing of interference signals. It considers the spatial distribution of interference signals and effectively separates and filters overlapping spectra in the frequency domain, thus obtaining the original signal and overcoming the limitations of traditional filtering techniques. The S300 uses an adaptive Kalman filter algorithm to process the original signal. Under severe interference, traditional methods are prone to waveform distortion and phase shifts. The adaptive Kalman filter algorithm can adaptively adjust according to the actual signal conditions, fully considering the impact of interference during signal reconstruction and restoring the true characteristics of the signal as much as possible, thereby obtaining a high-fidelity target signal and improving the accuracy of signal reconstruction. Based on high-fidelity target signals, the S400 constructs a multi-scale feature extraction and verification mechanism. Existing systems lack effective methods for quantifying signal quality and analyzing reliability, failing to provide reliable data quality assurance for subsequent systems. The multi-scale feature extraction and verification mechanism can extract and verify signal features at different scales, quantify signal quality, calculate signal reliability, and thus obtain a quantitatively purified signal with reliable metrics, providing reliable data quality assurance for subsequent systems. The S500 processes the quantitatively purified signal with reliable metrics based on a compensation strategy. Existing systems are mostly statically configured, lacking self-learning and self-optimization capabilities, making it difficult to adapt to dynamic changes in hydropower station operating conditions. The compensation strategy can dynamically compensate the signal according to changes in hydropower station operating conditions, further improving signal reliability and enabling the system to continuously improve its interference adaptability, obtaining a highly reliable output signal.

[0112] Through the comprehensive application of the above series of technical means, from interference source identification, filtering, signal reconstruction, quality assessment to signal compensation, a complete anti-interference processing flow is formed. Each step effectively solves the problems existing in the current technology, gradually removes the influence of interference signals, improves the quality and reliability of the signal, and the final high-reliability output signal is the result of multiple anti-interference processing, which can effectively resist various interferences in the complex environment of the hydropower station and ensure that the collected data is anti-interference data.

[0113] Secondly, this application discloses a hydropower station signal source side data anti-interference acquisition system, used to execute the hydropower station signal source side data anti-interference acquisition method disclosed in the first aspect of this application. The system specifically includes: The interference source distribution feature map acquisition module is used to construct a multi-source interference feature identification network based on the environmental conditions of the hydropower station and obtain the interference source distribution feature map. The raw signal acquisition module is used to obtain the raw signal based on the distribution feature map of the interference source through spatial-frequency domain joint filtering preprocessing; The high-fidelity target signal acquisition module is used to recursively process the original signal using an adaptive Kalman filter algorithm to obtain a high-fidelity target signal. The credible quantitative clean signal acquisition module is used to process high-fidelity target signals based on a multi-scale feature extraction and verification mechanism to obtain credible quantitative clean signals. The high-reliability output signal acquisition module is used to process the quantified clean signal based on the compensation strategy to obtain a high-reliability output signal.

[0114] Furthermore, the system disclosed in this application also includes: A multi-level redundant power supply architecture, including external dual-circuit power supply, internal UPS backup and hierarchical power supply management, is used to provide a stable and reliable power supply for signal acquisition equipment. The signal source side anti-interference protection system includes an electromagnetic shielding system, surge suppressor, and vibration-resistant design; The hot standby redundant signal processing control architecture includes a dual-CPU configuration, a redundant power supply configuration, and dual networks. Multiple signal quality protection loops, including both hardware and software-level dual protection mechanisms; Multi-protocol fusion signal transmission network and multi-path signal source information acquisition interface.

[0115] A multi-level redundant power supply architecture provides stable power to devices in the signal source-side anti-interference protection system. Devices and systems in the hot-standby redundant signal processing control architecture, such as dual-CPU configurations, redundant power supply configurations, and dual networks, have extremely high requirements for power stability. The multi-level redundant power supply architecture ensures reliable power to these critical devices under various conditions, preventing power failures from causing signal processing control architecture failure and thus affecting the operation of the entire data acquisition system. The hardware-level and software-level dual protection mechanisms in the multiple signal quality protection loops require stable power support. If hardware-level protection controllers and other devices fail due to power problems, they will be unable to protect and process signal quality in a timely manner. Software-level protection mechanisms also rely on stable power to maintain the normal operation of servers and other devices, enabling real-time monitoring and analysis of signal quality. Various communication devices, protocol conversion gateways, and fiber optic transmission systems in multi-protocol converged signal transmission networks all depend on a stable power supply. The multi-level redundant power supply architecture provides power assurance for these devices, ensuring stable and efficient signal transmission within the network.

[0116] Furthermore, the hot standby redundant signal processing control architecture specifically includes: 1) Deploying a dual-CPU hot standby main controller, which includes a main CPU and a backup CPU that synchronize data in real time via a high-speed PCIe 3.0 bus, with the switching cycle between the main CPU and the backup CPU not exceeding 2ms; 2) Constructing a triple redundant power supply system, which includes a primary main power supply, a secondary backup power supply, and a tertiary emergency power supply. The primary main power supply is supplied by a stable power source, the secondary backup power supply uses lithium batteries, and the tertiary emergency power supply includes a supercapacitor array. The primary main power supply comes directly from a stable power source, the secondary backup power supply uses a lithium battery UPS module, and the tertiary emergency power supply is equipped with a supercapacitor array. Seamless switching is achieved through an intelligent power management chip, ensuring that critical signal acquisition and data storage can still be maintained for 45 seconds under extreme power outage conditions; 3) Constructing a dual-ring redundant Ethernet topology architecture; the dual-ring redundant Ethernet topology architecture includes a main ring Ethernet topology architecture and a backup ring Ethernet topology architecture. The main ring Ethernet topology architecture uses cascaded gigabit Ethernet switches, and the backup ring Ethernet topology architecture uses an independent fiber optic loop.

[0117] Furthermore, the specific deployment methods for multiple signal quality protection loops include: 1) deploying a hardware-level signal quality protection controller based on an FPGA and ARM dual-core architecture; 2) constructing a real-time signal quality monitoring and analysis system and establishing a signal quality evaluation index system; 3) designing a software-level intelligent protection algorithm based on fuzzy logic control to achieve signal quality degradation trend prediction and fault mode identification; 4) establishing a five-level progressive signal quality protection strategy; 5) integrating a high-precision signal waveform recording and fault waveform recording integrated device and deploying a signal protection action log system; 6) establishing a comprehensive signal quality monitoring system and outputting protection strategies and abnormal trigger signals.

[0118] Furthermore, the construction method of the multi-protocol converged signal transmission network specifically includes: constructing a multi-protocol converged signal transmission network with a dual-star redundant topology; deploying an intelligent protocol conversion gateway cluster to achieve bidirectional transparent conversion between heterogeneous protocols; deploying a deterministic signal transmission mechanism based on IEEE 802.1Qbv time-sensitive networking; constructing a three-path redundant optical fiber signal transmission system to achieve automatic link switching within ≤10ms; establishing an end-to-end layered security encryption system and configuring hardware security modules to manage the key lifecycle; introducing a signal transmission integrity verification mechanism based on consortium blockchain to support signal data traceability and integrity auditing; and deploying an intelligent bandwidth management system to achieve intelligent allocation of network resources and congestion control.

[0119] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A hydropower station signal source side data anti-interference collection method, characterized in that, The method comprises the following steps: According to the environmental state of the hydropower station, a multi-source interference feature recognition network is constructed to obtain an interference source distribution feature map; Based on the interference source distribution feature map, the original signal is obtained through spatial-frequency domain joint filtering preprocessing; An adaptive Kalman filtering algorithm is used to recursively process the original signal to obtain a high-fidelity target signal; Based on the multi-scale feature extraction and verification mechanism, the high-fidelity target signal is processed to obtain a credibility quantized purified signal; Based on the compensation strategy, the credibility quantized purified signal is processed to obtain a high-reliability output signal.

2. The hydropower station signal source side data anti-interference collection method according to claim 1, characterized in that, The method comprises the following steps: Deploy a multi-dimensional sensor array network in the hydropower station and obtain global electromagnetic environment basic data; The global electromagnetic environment basic data is preprocessed, and multi-dimensional features are extracted from the preprocessed global electromagnetic environment basic data, including time domain features, frequency domain features, and time-frequency domain features; According to the historical normal operation data of the hydropower station equipment, the historical interference source information and the corresponding historical fault data, a labeled feature fingerprint library is constructed; Based on the labeled feature fingerprint library, a neural network is trained to obtain a deep learning interference source recognition model; According to the deep learning interference source recognition model, the multi-dimensional features are analyzed to obtain interference source positioning information; According to the interference source positioning information, a dynamic interference propagation modeling is established to obtain the interference influence domain distribution feature map.

3. The hydropower station signal source side data anti-interference collection method of claim 2, characterized in that, The method comprises the following steps: According to the interference source positioning information, boundary conditions and excitation sources are set to simulate the propagation of electromagnetic fields in the space of the hydropower station to obtain electromagnetic field propagation simulation, which includes electromagnetic field intensity distribution at different positions; A mechanical model of the hydropower station equipment is established, and the vibration propagation path of the equipment corresponding to the interference source positioning information is obtained through finite element analysis; An acoustic propagation model of the hydropower station equipment is established, and according to the interference source positioning information, the sound field distribution information is obtained, including sound pressure level distribution at different positions; According to the electromagnetic field propagation simulation, the vibration propagation path and the sound field distribution information, the linear superposition principle is used to add the interference intensities generated by different interference sources at the same position to obtain the superposition intensity distribution map of a single space point; According to the superposition intensity distribution map of all space points, the interference source distribution feature map is obtained.

4. The hydropower station signal source side data anti-interference collection method of claim 1, characterized in that, The method comprises the following steps: Spatial spectrum estimation is performed on the interference source distribution feature map to obtain a directional filtering signal; Based on the directional filtering signal, an adaptive frequency domain notch filter bank is constructed to obtain a frequency spectrum purification signal; Based on the frequency spectrum purification signal, multi-channel correlation analysis and blind source separation are performed to obtain a preliminary purified original signal.

5. The hydropower station signal source side data anti-interference collection method according to claim 4, characterized in that, Spatial spectrum estimation is performed on the interference source distribution feature map to obtain a directional filtering signal; The interference azimuth angle is extracted from the interference source distribution feature map; The weighting coefficients are applied to array receiving signals to obtain a directional filtering signal. The directional filtering signal is used to construct an adaptive frequency domain notch filter bank to obtain a spectrum purification signal, including:

6. The hydropower station signal source side data anti-interference collection method according to claim 5, characterized in that, The directional filtering signal is subjected to short-time Fourier transform to obtain a time-frequency graph; The time-frequency graph is used to identify the positions of interference spectral lines and their corresponding bandwidths; An adaptive notch filter bank is constructed according to the positions of the interference spectral lines and the bandwidths, the number of notch filters in the adaptive notch filter bank is consistent with the number of the positions of the interference spectral lines, and each notch filter corresponds to an interference signal; The time-frequency graph and the adaptive notch filter bank are multiplied point by point, and the multiplication result is subjected to inverse transform to obtain a spectrum purification signal. The spectrum purification signal is used to implement multi-channel correlation analysis and blind source separation to obtain a preliminarily purified original signal.

7. The hydropower station signal source side data anti-interference collection method according to claim 6, characterized in that, The spectrum purification signal is subjected to frame division to obtain a channel signal correlation matrix; A blind source separation model is established based on the channel signal correlation matrix; An iteration optimization separation matrix is obtained based on the blind source separation model when the separation meets the iteration convergence condition; A signal component that is most matched with a prior target feature is selected as a target component; The target component is subjected to amplitude normalization and phase correction to obtain a preliminarily purified original signal. The high-fidelity target signal is processed based on a multi-scale feature extraction and verification mechanism to obtain a credibility quantified purification signal, including:

8. The hydropower station signal source side data anti-interference collection method of claim 1, characterized in that, The high-fidelity target signal is subjected to multi-scale decomposition by using wavelet transform; Target features are extracted from the decomposed sub-signals of each scale, and the target features include one or more of energy features, variance features and mean features; The target features are subjected to feature anomaly verification, and the credibility weight of each sub-signal is determined according to the verification result; The sub-signals of each scale are combined by using the credibility weight to obtain a credibility quantified purification signal. The credibility quantified purification signal is processed based on a compensation strategy to obtain a high-reliability output signal, including:

9. The hydropower station signal source side data anti-interference collection method of claim 1, characterized in that, An abnormal value in the credibility quantified purification signal is identified by using a 3σ criterion; A corresponding compensation strategy is determined according to the type of the abnormal value, and is recorded as a target strategy; The target strategy is used for compensation to obtain a high-reliability output signal. The system comprises:

10. A hydropower station signal source side data anti-interference collection system, characterized in that, An interference source distribution feature map acquisition module, which is configured to construct a multi-source interference feature recognition network according to the environmental state of a hydropower station to obtain an interference source distribution feature map; An original signal acquisition module, which is configured to obtain an original signal by spatial-frequency domain joint filtering preprocessing based on the interference source distribution feature map; A high-fidelity target signal acquisition module, which is configured to obtain a high-fidelity target signal by recursively processing the original signal by using an adaptive Kalman filtering algorithm; A credibility quantified purification signal acquisition module, which is configured to obtain a credibility quantified purification signal by processing the high-fidelity target signal based on a multi-scale feature extraction and verification mechanism. ​ The high-reliability output signal acquisition module is configured to process the credibility quantification purification signal based on a compensation strategy to obtain a high-reliability output signal.

Images